INDEX

Explanations

An envelope

This neuron fires on runs of leading whitespace or indentation (long sequences of space/tab tokens at the start of a line).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 versuchen

-1.61

<i>

-1.52

 möchten

-1.52

あなたが

-1.50

of

-1.49

 cherchez

-1.49

 خبرگزاری

-1.45

 хора

-1.44

 sincere

-1.41

𝗻

-1.41

POSITIVE LOGITS

</em>

3.53

</i>

3.39

</strong>

3.25

</h3>

2.77

</h2>

1.98

similar

1.77

 manifestación

1.73

“

1.72

</u>

1.63

"))

1.63

Activations Density 0.044%